区间型符号数据的特征选择方法

对区间型符号数据进行特征选择,可以降低数据的维数,提取数据的关键特征。针对区间型符号数据的特征选择问题,本文提出了一种新的特征选择方法。首先,该方法使用区间数Hausdorff距离和区间数欧氏距离度量区间数的相似性,通过建立使得样本点与样本类中心相似性最大的优化模型来估计区间型符号数据的特征权重。其次,基于特征权重构建相应的分类器来评价所估计特征权重的优劣。最后,为了验证本文方法的有效性,分别在人工生成数据集和真实数据集上进行了数值实验,数值实验结果表明,本文方法可以有效地去除无关特征,识别出与类标号有关的特征。

A Feature Selection Method for Symbolic Interval Data

Feature selection for symbolic interval data can reduce the dimension of data and extract the key features of data.In order to deal with the feature selection problem, a new method is proposed in this paper. Firstly, Hausdorff distance and Euclidean distance are utilized to measure the similarity between two interval numbers, and an optimization model, which aims to maximize the similarity between each sample and its class center, is established to estimate the feature weights for symbolic interval data. Next, based on the estimated feature selection weights, a classifier is constructed to evaluate the goodness of the weights. Finally, in order to verify the effectiveness of the proposed method, numerical experiments are done in artificially generated data sets and real data sets, respectively. The numerical experiments results show that the proposed algrithm can eliminate irrelevant features and identify features which are relevant to the class labels.